Identifying Topic Shift and Topic Shading in Switchboard

Abstract

This paper highlights some of the ongoing work on the ADELE project, namely the identification and annotation of topic shift and topic shading in the Switchboard-1 Release-2 corpus. The purpose of this is to train an Artificial Neural Network to create a digital companion for the elderly that can communicate through informal,yet informed social dialogue, on a variety of topics of interest to a user over a prolonged time scale. To this end the project is focussing on topic shift and shading, the mechanisms which underpin the development of such conversations [6,8]. In the past, dialogue systems have predominantly focussed on practical tasks due to the complexity of modelling realistic everyday social talk [1]. With increasing awareness of the need for home robots and virtual home care agents to help assist in the provision of care for a rapidly ageing population, it is necessary to develop a more caring, involved, and personalised virtual care agent capable of such social dialogue.